215,756 Hits in 8.6 sec

A Learning Error Analysis for Structured Prediction with Approximate Inference

Yuanbin Wu, Man Lan, Shiliang Sun, Qi Zhang, Xuanjing Huang
2017 Neural Information Processing Systems  
The error analyses also suggest a new margin for existing learning algorithms.  ...  In this work, we try to understand the differences between exact and approximate inference algorithms in structured prediction.  ...  Yuanbin Wu is supported by a Microsoft Research Asia Collaborative Research Program.  ... 
dblp:conf/nips/WuLSZH17 fatcat:lmn7ewiaeje4nptg2fg46ltwqq

A bias/variance decomposition for models using collective inference

Jennifer Neville, David Jensen
2008 Machine Learning  
Conventional analysis decomposes loss into errors due to aspects of the learning process, but in relational domains, the inference process used for prediction introduces an additional source of error.  ...  Bias/variance analysis is a useful tool for investigating the performance of machine learning algorithms.  ...  Acknowledgements We thank our anonymous reviewers and Cindy Loiselle for their thoughtful and constructive comments.  ... 
doi:10.1007/s10994-008-5066-6 fatcat:vkyukzo2v5c3dc42ewqgqb3syu

Deep Component Analysis via Alternating Direction Neural Networks [article]

Calvin Murdock, Ming-Fang Chang, Simon Lucey
2018 arXiv   pre-print
By interpreting feed-forward networks as single-iteration approximations of inference in our model, we provide both a novel theoretical perspective for understanding them and a practical technique for  ...  On the other hand, shallow representation learning with component analysis is associated with rich intuition and theory, but smaller capacity often limits its usefulness.  ...  Component Analysis and Matrix Factorization Component analysis is a common approach for shallow representation learning that approximately decomposes data x ∈ R d into linear combinations of learned components  ... 
arXiv:1803.06407v1 fatcat:tfivbuxbvbfc5lepgeglb5gpru

How Hard is Inference for Structured Prediction?

Amir Globerson, Tim Roughgarden, David A. Sontag, Cafer Yildirim
2015 International Conference on Machine Learning  
The goal of this paper is to develop a theoretical explanation of the empirical effectiveness of heuristic inference algorithms for solving such structured prediction problems.  ...  Structured prediction tasks in machine learning involve the simultaneous prediction of multiple labels.  ...  Taken together, our results provide the first theoretical analysis for structured prediction using approximate inference.  ... 
dblp:conf/icml/GlobersonRSY15 fatcat:obo3xnryfzbvdfhd3mby4ktjnu

An analysis of how ensembles of collective classifiers improve predictions in graphs

Hoda Eldardiry, Jennifer Neville
2012 Proceedings of the 21st ACM international conference on Information and knowledge management - CIKM '12  
We present a theoretical analysis framework that shows how ensembles of collective classifiers can improve predictions for graph data.  ...  We show how collective ensemble classification reduces errors due to variance in learning and more interestingly inference.  ...  Government is authorized to reproduce and distribute reprints for governmental purposes notwithstanding any copyright notation hereon.  ... 
doi:10.1145/2396761.2396793 dblp:conf/cikm/EldardiryN12 fatcat:7ac3s5usuzf7lfyka45antz2w4

Conditional Random Fields for Pattern Recognition Applied to Structured Data

Tom Burr, Alexei Skurikhin
2015 Algorithms  
Therefore, conditional random fields (CRFs) model structured data using the conditional distribution P(Y|X = x), without specifying a model for P(X), and are well suited for applications with dependent  ...  Image analysis is one setting where one might want to infer whether a pixel patch contains an object that is "manmade" (such as a building) or "natural" (such as a tree).  ...  Acknowledgments We acknowledge Los Alamos National Laboratory's directed research and development (LDRD) for funding this work. Author Contributions Both authors contributed equally to this work.  ... 
doi:10.3390/a8030466 fatcat:ylji3stlyzcdlc6yjqcfpklk7e

Ensemble Learning for Relational Data

Hoda Eldardiry, Jennifer Neville, Ryan A. Rossi
2020 Journal of machine learning research  
We show that ensembles of collective classifiers can improve predictions for graph data by reducing errors due to variance in both learning and inference.  ...  In addition, we propose a relational ensemble framework that combines a relational ensemble learning approach with a relational ensemble inference approach for collective classification.  ...  Acknowledgments We thank Luc De Raedt for many insightful suggestions and feedback that greatly improved the manuscript. We also thank all the reviewers for many helpful suggestions and feedback.  ... 
dblp:journals/jmlr/EldardiryNR20 fatcat:ai42eks7zbf7fi3tco4k3cjw54

Training structural SVMs when exact inference is intractable

Thomas Finley, Thorsten Joachims
2008 Proceedings of the 25th international conference on Machine learning - ICML '08  
We provide a theoretical and empirical analysis of both types of approximate trained structural SVMs, focusing on fully connected pairwise Markov random fields.  ...  This leads to a need for approximate training methods. Unfortunately, knowledge about how to perform efficient and effective approximate training is limited.  ...  Acknowledgments This work was supported under NSF Award IIS-0713483 and through a gift from Yahoo! Inc.  ... 
doi:10.1145/1390156.1390195 dblp:conf/icml/FinleyJ08 fatcat:cktss4iulzfaroqx3mdj2xj47y

Efficiently Learning Random Fields for Stereo Vision with Sparse Message Passing [chapter]

Jerod J. Weinman, Lam Tran, Christopher J. Pal
2008 Lecture Notes in Computer Science  
We present a novel CRF for stereo matching with an explicit occlusion model and propose sparse message passing to dramatically accelerate the approximate inference needed for parameter optimization.  ...  As richer models for stereo vision are constructed, there is a growing interest in learning model parameters.  ...  [10] explored sparse methods for approximate inference using BP in chain-structured graphs.  ... 
doi:10.1007/978-3-540-88682-2_47 fatcat:aw4u5uwrjvh6njjpgcuop3gpqi

On Learning Conditional Random Fields for Stereo

Christopher J. Pal, Jerod J. Weinman, Lam C. Tran, Daniel Scharstein
2010 International Journal of Computer Vision  
We explore a number of novel CRF model structures including a CRF for stereo matching with an explicit occlusion model.  ...  To accelerate approximate inference we have developed a new method called sparse variational message passing which can reduce inference time by an order of magnitude with negligible loss in quality.  ...  Acknowledgements We would like to thank Anna Blasiak and Jeff Wehrwein for their help in creating the data sets used in this paper.  ... 
doi:10.1007/s11263-010-0385-z fatcat:5vbs4bup2rc3jkp5bxsbniq5ti

Active Learning of Spin Network Models [article]

Jialong Jiang, David A. Sivak, Matt Thomson
2019 arXiv   pre-print
Therefore, we propose a general mathematical framework to study inference with iteratively applied perturbations.  ...  Our active learning framework could be powerful in the analysis of complex networks as well as in the rational design of experiments.  ...  The mean interaction strength prediction error and structural edge prediction accuracy, as a function of sample size, are shown in Fig. S5 .  ... 
arXiv:1903.10474v3 fatcat:vbbztyeo4vhqhechyfb7ncdmo4

Learning and Inference for Structured Prediction: A Unifying Perspective

Aryan Deshwal, Janardhan Rao Doppa, Dan Roth
2019 Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence  
In a structured prediction problem, one needs to learn a predictor that, given a structured input, produces a structured object, such as a sequence, tree, or clustering output.  ...  Prototypical structured prediction tasks include part-of-speech tagging (predicting POS tag sequence for an input sentence) and semantic segmentation of images (predicting semantic labels for pixels of  ...  Acknowledgements This work was supported in part by NSF grants #1910213 and #1845922, and in part by Contracts W911NF15-1-0461 and HR0011-17-S-0048 with the US Defense Advanced Research Projects Agency  ... 
doi:10.24963/ijcai.2019/878 dblp:conf/ijcai/DeshwalDR19 fatcat:lmpr57bvajfqdmafbkdzg4dpi4

The local paradigm for modeling and control: from neuro-fuzzy to lazy learning

Gianluca Bontempi, Hugues Bersini, Mauro Birattari
2001 Fuzzy sets and systems (Print)  
This paper presents a comparative analysis of two di erent local approaches: the neuro-fuzzy inference system and the lazy learning approach.  ...  Neuro-fuzzy is a hybrid representation which combines the linguistic description typical of fuzzy inference systems, with learning procedures inspired by neural networks.  ...  It is worth noting that this also means that lazy learning can return, along with each predicted value, an estimation of its standard error.  ... 
doi:10.1016/s0165-0114(99)00172-4 fatcat:6w2rggp2tndgvpwg5apf4qsnaa

SVM learning of IP address structure for latency prediction

Robert Beverly, Karen Sollins, Arthur Berger
2006 Proceedings of the 2006 SIGCOMM workshop on Mining network data - MineNet '06  
SVM regression on a large, randomly collected data set of 30,000 Internet latencies yields a mean prediction error of 25ms using only 20% of the samples for training.  ...  Our results are promising for equipping end-nodes with intelligence for service selection, user-directed routing, resource scheduling and network inference.  ...  Acknowledgments We thank Mike Afergan, Steve Bauer, Srikanth Kandula and our reviewers for their invaluable comments and suggestions.  ... 
doi:10.1145/1162678.1162682 dblp:conf/minenet/BeverlySB06 fatcat:wmos63722rh75pxofubyilrsf4

Guest editors' introduction: special issue on inductive logic programming (ILP-2007)

Hendrik Blockeel, Jude Shavlik, Prasad Tadepalli
2008 Machine Learning  
They perform empirical studies that demonstrate that inference can be a substantial contributor to errors in prediction and also provide insight into common approaches to statistical-relational learning  ...  In their paper Neville and Jensen extend the classic bias-variance analysis to account for the impact of approximation during the inference process that commonly accompanies statistical models, especially  ... 
doi:10.1007/s10994-008-5078-2 fatcat:xqqvh4peh5bi3fmtmc7je5uu7a
« Previous Showing results 1 — 15 out of 215,756 results